This case requires us to address a claim for age discrimination in disability benefits in light of the Supreme Court’s decision in Hazen Paper Co. v. Biggins, 507 U.S. 604, 113 S.Ct. 1701, 123 L.Ed.2d 338 (1993).
Two individuals are hired as police officers on the same day. One is twenty-five, the other forty-five. A year later, they are injured in the same accident. Under the California Public Employees’ Retirement Law, Cal. Gov’t Code § 21417, the younger employee receives 50% of final monthly compensation as a disability benefit, but the older employee receives only 20% the difference due solely to their ages at hire. Does a challenge to such a plan state a claim for disparate treatment or disparate impact under the Age Discrimination in Employment Act (“ADEA”), 29 U.S.C. §§ 621-634? We hold that it does, and reverse the district court’s grant of motions to dismiss for failure to state a claim and for judgment on the pleadings.
BACKGROUND
Appellants, former police officers, correctional officers and other “safety employees”2 of the State of California and local agencies (collectively, the “Employees”), brought this class action to challenge calculation of disability benefits under the California Public Employees Retirement System (“PERS,” also known as “Cal-PERS”). The Employees were all hired at age 40 or later and retired from their jobs because of industrial disabilities. Appel-lees are PERS, the State of California and three local cities and a county which contracted for benefits under PERS (collectively, “PERS”). Employees of state agencies are automatically covered by PERS; local agencies may join PERS to provide benefits for their employees. Cal. Gov’t Code §§ 20281, 20300, 20460.
Benefits Under PERS
PERS provides two types of benefits: service retirement benefits for employees who work until a normal retirement age, and industrial disability retirement benefits for employees who are injured on the job. and retire with a disability. Under the PERS system, the level of compensation at the time of retirement and actual years of service are the key factors in calculation of service benefits. For example, a safety employee who works until normal retirement age receives benefits calculated as follows: *693See, e.g., Cal. Gov’t Code §§ 21362, 21363. Thus, an employee who worked twenty-five years before retiring would receive 50% of final monthly compensation as service retirement benefits (2% x 25 years), whereas an employee who retired after working 13 years would receive 26% of final compensation (2% x 13 years).
Disability retirement benefits are calculated according to a different formula from service benefits. Until 1980, an employee’s final compensation was the only variable in computing disability retirement benefits. Under California Government Code section 21292.5,4 any safety employee injured on the job received 50% of final compensation as a disability benefit, regardless of age or the number of years of actual service:
Upon retirement of a state safety member for industrial disability he or she shall receive a disability retirement allowance of 50 percent of his or her final compensation....
In 1980, the California Legislature enacted section 21292.6 (now recodified as section 21417) to provide a limitation on this flat percentage:
Notwithstanding any -other provision of this part, the industrial disability retirement allowance of a member ... shall not exceed the service retirement allowance that would be payable as a result of service ... if the member’s service had continued to age 55, if [the employee is] a patrol, state peace officer/firefighter, state safety, or local safety member....
Cal. Gov’t Code § 21417.
Under the new statutory scheme, a safety employee injured on the job receives the lesser of (1) 50% of final compensation, or (2) the amount that employee would have received in service retirement benefits had the employee continued to work until the age of 55. In order to derive this latter amount, section 21417 imposes a statutory ceiling of 55 — a presumed, but not mandatory retirement age under California law — and calculates an employee’s “potential years of service” by subtracting the employee’s age at hire from 55. The final compensation and “years of potential service” are the key variables in the calculation. The formula can be expressed as follows:
monthly disability benefits = 2% x monthly salary x (55 — age at hire)
If the “potential years of service” calculation is less than 50%, the employee will receive only the lower amount as a disability retirement benefit, not the otherwise available 50%.5
In mathematical terms, then, section 21417 limits disability .retirement benefits to the lesser of 50% or 2% x (55 — age at hire). The only variable that matters in the calculation is the employee’s age at hire.
District Court Proceedings
In their First Amended Complaint, the Employees alleged that section 21417 violates the ADEA under a disparate treatment theory because it reduces disability benefits based solely upon workers’ ages at hire. The district court disagreed and granted the motions to dismiss and for judgment on the pleadings, holding that the Employees failed to state a disparate treatment claim under Hazen. The court, granted leave to amend the complaint to allege a disparate impact claim. The district court then dismissed that claim, finding that ADEA regulations authorized the *694statutory scheme in section 21417, and alternatively, that the benefit differential was justified by “business necessity.”
STANDARD OF REVIEW
Because this case comes to us after dismissal pursuant to Federal Rules of Civil Procedure 12(b)(6) and 12(c), we review the district court’s decisions de novo. See Steckman v. Hart Brewing, Inc., 143 F.3d 1293, 1295 (9th Cir.1998) (dismissal for failure to state a claim pursuant to Fed.R.Civ.P. 12(b)(6)); Nelson v. City of Irvine, 143 F.3d 1196, 1200 (9th Cir.) (judgment on the pleadings pursuant to Fed.R.Civ.P. 12(c)), cert. denied, — U.S. -, 119 S.Ct. 444, 142 L.Ed.2d 399 (1998).
DISCUSSION
The ADEA prohibits employers from discriminating “against any individual with respect to his compensation, terms, conditions, or privileges of employment because of such individual’s age.” 29 U.S.C. § 623(a)(1). As the Supreme Court has noted, “[t]he Courts of Appeals repeatedly have faced the question whether an employer violates the ADEA by acting on the basis of a factor ... that is empirically correlated with age.” Hazen, 507 U.S. at 608, 113 S.Ct. 1701. Here we consider whether “potential years of service” under section 21417 is equivalent to “actual years of service,” which, according to Hazen, is merely empirically correlated with age and does not violate the ADEA.
I. Disparate Treatment
The parties dispute the significance of Hazen. The Employees claim that it is irrelevant. This position underestimates the teaching of Hazen. Conversely, PERS argues that Hazen is dispositive, because “potential years of service” under section 21417 = is merely a permissible variant of the “actual years of service” standard countenanced by the Supreme Court. Hazen, however, does not extend so far.
In Hazen, the employer terminated a 62-year old employee who was a few weeks shy of having his pension vest. The Supreme Court held that firing an employee “solely because he has nine-plus years of service and therefore is ‘close to vesting’ would not constitute discriminatory treatment on the basis of age,” regardless of the fact that older employees are more likely to be “close to vesting” than their younger counterparts. Id. at 611-12, 113 S.Ct. 1701. When an employer acts on the basis of years of service, a factor that is merely “empirically correlated” with age, id. at 608, 113 S.Ct. 1701, a disparate treatment claim must fail:
When the employer’s decision is wholly motivated by factors other than age, the problem of inaccurate and stigmatizing stereotypes disappears. This is true even if the motivating factor is correlated with age....
Id. at 611, 113 S.Ct. 1701 (emphasis in original). The Court went on to explain that age is “analytically distinct” from years of service because a younger employee “may have worked for a particular employer his entire career, while an older worker may have been newly hired.”. Id.
In exempting employment decisions based on actual years of service from liability under the ADEA, however, the Court expressly stated that it was not addressing the “special case” where an employee’s pension vests “as a result of his age....” Id. at 613, 113 S.Ct. 1701 (emphasis added). We are now squarely presented with the “special case” left undecided in Hazen—a plan under which employment benefits differ solely “as a result of’ age. Whereas the employer in Hazen did not violate the ADEA because a decision based on actual years of service permits the employer to “ignore an employee’s age,” see id. at 612, 113 S.Ct. 1701, the calculation of “potential years of service” under section 21417 is entirely a function of the employee’s age at hire. See Cal. Gov’t Code § 21417 (calculating *695disability benefits at 2% x (55 — age at hire)).6
The net effect of section 21417 can be illustrated as follows:
Disability Benefit Employee’s Age Prior to Section at Hire 21417 Disability Benefit Under Section 21417
30 50% 50%
35 50% 40%
40 50% 30%
45 50% 20%
50 50% 10%
The question, then, is whether “potential years of service” under section 21417 is analytically equivalent to “actual years of service” under Hazen. We hold that it is not. In asserting that “potential years of service” for disability retirement benefits is merely a derivative of the “actual years of service” standard for service retirement benefits, PERS ignores one critical fact: while actual years of service may tend to be “empirically correlated” with age, Hazen, 507 U.S. at 608, 113 S.Ct. 1701, the calculation of “potential years of service” under section 21417 is a strict function of age. It bears no relationship to length of service, seniority or other factor that might permissibly be “empirically correlated” with age. ’ Were we to credit PERS’ argument that section 21417 simply equates disability retirement benefits with service retirement benefits, we would be elevating form over substance. The practical application of section 21417 leaves no doubt that age at hire, not actual years worked or any variant thereof, is the sole basis for lower benefits.
Section 21417 thus represents the photographic negative of Hazen, which does not immunize PERS’ age-based benefits plan. Hazen teaches that disparate treatment occurs when “ ‘the employer treats some people less favorably than others because of race, color, religion [or other protected characteristics].’ ” Hazen, 507 U.S. at 609, 113 S.Ct. 1701 (quoting International Bhd. of Teamsters v. United States, 431 U.S. 324, 335 n. 15, 97 S.Ct. 1843, 52 L.Ed.2d 396 (1977)) (bracketed material in original). This circuit has adopted a “but for” test in analyzing disparate treatment claims under the ADEA. See E.E.O.C. v. Local 350, Plumbers and Pipefitters, 998 F.2d 641, 646 (9th Cir.1992) (“Local 350”) (“[A]n employer discriminates ‘because of age [in violation of ADEA] whenever age is a ‘but for’ cause of discrimination.”) (quoting E.E.O.C. v. Borden’s, Inc., 724 F.2d 1390, 1393 (9th Cir.1984)).7 It is unquestionable that the Employees would have received greater disability retirement benefits but for their older ages at hire. Id.; see also Huff v. UARCO, Inc., 122 F.3d 374, 388 (7th Cir.1997) (policy violates ADEA where “age is an express condition of receiving a benefit,” and “[b]ut for the age-related restriction,” certain older workers would have received a benefit available to other employees). Under Hazen and Local 350, the Employees stated a disparate treatment claim.
We next address PERS’ additional arguments supporting dismissal. PERS contends that the Employees cannot complain about lower disability benefits because they “voluntarily” chose to begin work at a later age. This argument not only ignores the reality of the modern workforce, it conflicts with our rejection of precisely the same argument in Local 350. See Local *696350, 998 F.2d at- 646 (rejecting the argument that a policy which discriminated against retirees did not violate the ADEA because those individuals voluntarily decided to retire). PERS further claims that the system presumes that the younger employee, having been in the workforce less time than an older hire, does not have other resources to count on, “unlike an employee who begins work in the public sector later in life.” Not only is this wholly speculative, since many employees first enter the workforce later in life, it epitomizes precisely the sort of class-based assumptions that the Supreme Court has repeatedly rejected. See, e.g., Los Angeles Dept. of Water and Power v. Manhart, 435 U.S. 702, 98 S.Ct. 1370, 55 L.Ed.2d 657 (1978) (employer violates Title VII’s prohibition on discrimination against individuals where employer’s retirement contribution scheme assumes that women generally live longer than men).
PERS finally argues that the Employees failed to state a claim because section 21417 was enacted for economic, not discriminatory reasons. Guided by Hazen, we note that “[p]roof of discriminatory motive is critical, although it can in some situations be inferred from the mere fact of differences in treatment .... In a disparate treatment case, liability depends on whether the protected trait ... actually motivated the employer’s decision.” 507 U.S. at 609, 610 (internal quotation marks and citations omitted). We decline to make ^ny such determination on appeal from a 12(b)(6) dismissal. Both sides invoked the legislative history to support their position. For example, the Employees offered a memorandum from the legislative bill file of the Senate Public Employees and Retirement Committee which explained that “[t]he members affected would be those first [hired] at advanced ages,” as well as a letter from the California State Employees Association to the sponsor of section 21417 commenting on the perceived unfairness of section 21417: “[A] person who becomes industrially disabled ... would have the same needs, regardless of age.”8 PERS countered with a Legislative Analyst’s report to show that section 21417 was intended to be a cost-saving measure: “[T]his bill is expected to reduce future PERS program retirement costs.” Although the Employees have met the threshold pleading requirements for a disparate treatment claim, the necessity of establishing a discriminatory motive highlights the factual nature of the dispute, making the case inappropriate for dismissal at this early stage. Cf. Village of Arlington Heights v. Metropolitan Hous. Dev. Corp., 429 U.S. 252, 266, 97 S.Ct. 555, 50 L.Ed.2d 450 (1977) (determining whether “discriminatory purpose was a motivating factor demands a sensitive inquiry into such circumstantial and direct evidence of intent as may be available”).
II. Disparate Impact
The district court also erred in dismissing the Employees’ disparate impact claim in reliance on “applicable age discrimination regulations”9 and on grounds of “business necessity.” A threshold question is whether disparate impact theory survives Hazen. In Local 350, we explicitly held that disparate impact claims re*697main cognizable under the ADEA after Hazen:
We perceive no conflict between Hazen and our decision in this case.... [I]n this circuit a plaintiff may challenge age discrimination under a disparate impact analysis....
998 F.2d at 648 n.2 (as amended on reconsideration). While we recognize the Supreme Court’s intimation that a disparate impact claim may not be viable under the ADEA,10 we are bound, PERS’ arguments notwithstanding, by Local 350’s express reaffirmation of disparate impact theory. See In re Visness, 57 F.3d 775, 778 (9th Cir.1995).
Having alleged that the consequences of section 21417 fall more harshly on “those employees hired at age 40 and over,” and documenting the impact of the benefit calculations,11 the Employees adequately stated a disparate impact claim. See Hazen, 507 U.S. at 609, 113 S.Ct. 1701; see also Pfaff v. United States Dept. of Hous. and Urban Dev., 88 F.3d 739, 745 (9th Cir.1996) (prima facie case of discrimination under a disparate impact theory requires plaintiff to show “ '(1) the occurrence of certain outwardly neutral ... practices, and (2) a significantly adverse or disproportionate impact on persons of a particular [type] produced by the [defendant’s] facially neutral acts or practices’ ”) (bracketed material in original) (quoting Palmer v. United States, 794 F.2d 534, 538 (9th Cir.1986)); Local 350, 998 F.2d at 648 n. 2 (EEOC’s challenge to a policy which adversely affected only retirees is cognizable under a disparate impact theory).
It is true that a disparate impact claim fails where a practice can “be justified by business necessity,” see Hazen, 507 U.S. at 609, 113 S.Ct. 1701, and that PERS defends section 21417 as an effort to save costs. PERS argues that section 21417 was directed at reducing the high cost of providing disability retirement benefits at a rate of 50% of final compensation to employees who, because they began working later in life, would likely never have accumulated enough “actual years of service” to reach that amount in service retirement benefits. On the record before us, whether such a rationale constitutes a sufficient “business necessity” cannot be determined on a motion to dismiss, and the district court erred in dismissing the Employees’ ■ disparate impact claim at the pleading stage. Cf. Gilligan v. Jamco Dev. Corp., 108 F.3d 246, 250-51 (9th Cir.1997) (on the record before the court, business necessity defense not appropriate basis for dismissal of disparate impact claim at pleading stage).
REVERSED AND REMANDED.
. "Safety employees” is an employment classification that includes, among others, patrol officers, and state peace officers and firefighters. Cal. Gov't Code §§ 20371(b), 20403. monthly service benefits = 2% % monthly salary * actual years of service x statutory multiplier3
.The multiplier ranges from 1.0 to 1.35, depending on the employee's age at retirement, and, for safety employees, does not increase after age 55. Because the parties agree that the effect of the multiplier is irrelevant to this appeal, we omit it from further consideration and discussion.
. The statute was recodified as section 21411 in 1995.
. For example, if an employee were hired at 38 and later injured on the job, section 21417 would presume 17 “potential years of service” (55-38). Because 2% x 17 = 34%, which is less than the statutory default amount of 50%, the injured employee would receive only 34% of final compensation as a disability retirement benefit. An employee hired at 42 and injured on the job would receive only 26%. Successive calculations demonstrate that it is the age at hire, not actual years of service or any other variable, that is determinative of the benefit calculation.
. The use of 55 as an artificial ceiling further exacerbates the disparity between the benefits provided to younger and older employees. Under PERS' current scheme, a safety employee hired at the age of 45 but injured one year later would receive only 20% of final compensation as a disability retirement benefit [2% x (55-45) ]. If the ceiling were 65, the mandatory retirement age for state safety members, see Cal. Gov't Code § 21132, the employee would instead receive 40% of final compensation as a disability retirement benefit [2% x (65-45) ]. Of course, an employee hired before the age of 30 would always receive 50% of final compensation as a disability benefit regardless of the ceiling.
. Because the Sixth Circuit does not follow this circuit’s "but for” test of discrimination, PERS' reliance on Lyon v. Ohio Educ. Ass'n and Prof'l Staff Union, 53 F.3d 135 (6th Cir.1995), is misplaced. Lyon is also distinguishable on its facts. See infra at n. 8.
. While such documents might support the Employees’ position that the Legislature knew of and intended the discriminatory effects, see Borden's, 724 F.2d at 1393 ("[T]he discrimination was intentional in the sense that Borden purposefully drafted its ... policy to have this effect. We need look no further for the intent necessary to support a finding of discrimination under the disparate treatment theory.”), we do not decide that issue here. We note, however, that this evidence from the legislative history distinguishes this case from Lyon. See Lyon, 53 F.3d at 139 (noting that ”[t]here was no evidence that either defen-danl intended ... to discriminate against older employees”).
. The district court did not cite any specific regulations, but the only relevant one is inapplicable here. Although an employer is entitled to distinguish among employees by years of service as long as that distinction is made "without regard to age,” see 29 U.S.C. § 623(f)(2); see also Quinones v. City of Evanston, 58 F.3d 275, 278-79 (7th Cir.1995), that regulation does not defeat the Employees' claim because section 21417 limits benefits on the sole basis of age.
. See Hazen, 507 U.S. at 610, 113 S.Ct. 1701 ("[W]e have never decided whether a disparate impact theory of liability is available under the ADEA.... Disparate treatment ... captures the essence of what Congress sought to prohibit in the ADEA.”). See also, e.g., Mullin v. Raytheon Co., 164 F.3d 696, 703 (1st Cir.1999) (holding that disparate impact claims are not cognizable under the ADEA); Ellis v. United Airlines, Inc., 73 F.3d 999, 1007 (10th Cir.1996) (same); E.E.O.C. v. Francis W. Parker School, 41 F.3d 1073, 1076-7.8 (7th Cir.1994) (same).
. The district court suggested that section 21417 actually adversely impacts younger employees, because younger disabled employees would have received far more than 50% of their salary as disability benefits in absence of the limitation in section 21417. This is incorrect. The default amount for disabled employees is 50%, unless section 21417 applies to limit the employee to an amount lower than 50%. See, e.g., Cal. Gov’t Code §§ 21411, 21417. A younger employee would therefore always receive the default amount of 50%, no more, and would be unaffected by section 21417’s limitation.